Project 4

---

by Jerry Catron

# Importing Libraries
import folium
import pandas as pd
import geopandas as gpd
import numpy as np
import warnings

# Getting the data
arrest_table = pd.read_csv("https://cmsc320.github.io/files/BPD_Arrests.csv")

arrest_table = arrest_table[pd.notnull(arrest_table["Location 1"])]

arrest_table["lat"], arrest_table["long"] = arrest_table["Location 1"].str.split(",").str
arrest_table["lat"] = arrest_table["lat"].str.replace("(", "").astype(float)
arrest_table["long"] = arrest_table["long"].str.replace(")", "").astype(float)

# surpressing deprecation warning
warnings.filterwarnings('ignore');

arrest_table.head()

# Aggregating the # of arrests per neighborhood
neighborhoods_counts = arrest_table.groupby(['neighborhood']).count()
# reseting index so that the neighborhood names are a column instead of the index variable
neighborhood_arrest_counts = neighborhoods_counts.loc[:,'arrest'].reset_index()
#renaming arrest column to arrest count
neighborhood_arrest_counts = neighborhood_arrest_counts.rename(columns= {"arrest":"arrest_count"})
neighborhood_arrest_counts.head()

	neighborhood	arrest_count
0	Abell	58
1	Allendale	296
2	Arcadia	75
3	Arlington	682
4	Armistead Gardens	142

# check how many rows we have
neighborhood_arrest_counts.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 267 entries, 0 to 266
Data columns (total 2 columns):
 #   Column        Non-Null Count  Dtype 
---  ------        --------------  ----- 
 0   neighborhood  267 non-null    object
 1   arrest_count  267 non-null    int64 
dtypes: int64(1), object(1)
memory usage: 4.3+ KB

# Getting Baltimore's neighborhood geodata

# GeoJSON source
neighborhood_geo_url = 'https://opendata.arcgis.com/datasets/f63445d9528349febb81f0b89cd8be5b_0.geojson'

# Using geopandas to read the geoJSON data into a Dataframe, converting shape values into POLYGON shapes for each neighborhood with coordinates
geoJSON_df = gpd.read_file(neighborhood_geo_url)

# Only need the neighborhood names, population density, and the geometry of each neighborhood
geoJSON_df = geoJSON_df.loc[:,['name','geometry']]

geoJSON_df.head()

	name	geometry
0	Abell	POLYGON ((-76.61113 39.32345, -76.61060 39.323...
1	Allendale	POLYGON ((-76.67263 39.29184, -76.67330 39.291...
2	Arcadia	POLYGON ((-76.56853 39.33595, -76.56859 39.336...
3	Arlington	POLYGON ((-76.68627 39.34791, -76.68601 39.347...
4	Armistead Gardens	POLYGON ((-76.55880 39.30646, -76.55860 39.306...

# check how many rows we have
geoJSON_df.info()

<class 'geopandas.geodataframe.GeoDataFrame'>
RangeIndex: 278 entries, 0 to 277
Data columns (total 2 columns):
 #   Column    Non-Null Count  Dtype   
---  ------    --------------  -----   
 0   name      278 non-null    object  
 1   geometry  278 non-null    geometry
dtypes: geometry(1), object(1)
memory usage: 4.5+ KB

As we can see the GeoJSON data has more neighborhoods than our arrest data. So let's see what's different

missing_hoods = np.setdiff1d(neighborhood_arrest_counts.neighborhood.values, geoJSON_df.name.values)

print(missing_hoods)

['Baltimore-Linwood' 'Booth-Boyd' 'Chinquapin Park-Belvedere'
 'Christopher' 'Claremont-Freedom' 'Glenham-Belford'
 'Harford-Echodale/Perring Parkway' 'Joseph Lee' 'SBIC'
 'Washington Village' 'Wilson Heights' 'Woodring']

After some manual inspection of each table, I found some pairs that had a different name for the same area. However there are still some remaining neighborhoods there is no data for, so we will not include those on our map.

# Changing neighborhood names to match data in the two tables
neighborhood_arrest_counts.replace({'Booth-Boyd': 'Boyd-Booth', 'Chinquapin Park-Belvedere': 'Belvedere', 'SBIC': 'South Baltimore'}, inplace=True)

geoJSON_df.replace({'Glenham-Belhar': 'Glenham-Belford', 'Washington Village/Pigtown': 'Washington Village'}, inplace=True)

# Renaming column so there is a key value we can merge on
geoJSON_df = geoJSON_df.rename(columns= {"name":"neighborhood"})

# Merging the arrest 
merged_df = geoJSON_df.merge(neighborhood_arrest_counts, on = 'neighborhood')
merged_df

	neighborhood	geometry	arrest_count
0	Abell	POLYGON ((-76.61113 39.32345, -76.61060 39.323...	58
1	Allendale	POLYGON ((-76.67263 39.29184, -76.67330 39.291...	296
2	Arcadia	POLYGON ((-76.56853 39.33595, -76.56859 39.336...	75
3	Arlington	POLYGON ((-76.68627 39.34791, -76.68601 39.347...	682
4	Armistead Gardens	POLYGON ((-76.55880 39.30646, -76.55860 39.306...	142
...	...	...	...
255	Yale Heights	POLYGON ((-76.69292 39.27249, -76.69282 39.272...	81
256	Belvedere	POLYGON ((-76.60270 39.35934, -76.60263 39.359...	32
257	Belair-Edison	POLYGON ((-76.57629 39.32211, -76.57579 39.321...	1485
258	Four By Four	POLYGON ((-76.57995 39.31617, -76.58027 39.315...	102
259	Charles Village	POLYGON ((-76.61539 39.31791, -76.61539 39.317...	367

260 rows × 3 columns

Much of the following code for formatting this Folium map is sourced from Navid Mashinchi's article "How to Step Up your Folium Choropleth Map skills". Source

# Plotting a map of Baltimore using folium
map_osm = folium.Map(location=[39.29, -76.61], zoom_start=11)

# Adding choropleth "heat map" of arrest count by neighborhood
folium.Choropleth(
    geo_data=merged_df,
    data=merged_df,
    columns=['neighborhood', 'arrest_count'],
    key_on='feature.properties.neighborhood',
    fill_color='YlOrRd',
    fill_opacity=0.6,
    line_opacity=0.2,
    legend_name="arrests",
    smooth_factor=0,
    Highlight= True,
    line_color = "#414a4c",
    name = "Arrests by Baltimore Neighborhood",
    show=True,
    overlay=True,
    nan_fill_color = "White"
).add_to(map_osm)

# Add hover functionality.
style_function = lambda x: {'fillColor': '#ffffff', 
                            'color':'#000000', 
                            'fillOpacity': 0.1, 
                            'weight': 0.1}
highlight_function = lambda x: {'fillColor': '#000000', 
                                'color':'#000000', 
                                'fillOpacity': 0.50, 
                                'weight': 0.1}
HOV = folium.features.GeoJson(
    data = merged_df,
    style_function=style_function, 
    control=False,
    highlight_function=highlight_function, 
    tooltip=folium.features.GeoJsonTooltip(
        fields=['neighborhood','arrest_count'],
        aliases=['Neighborhood','Arrest Count'],
        style=("background-color: white; color: #333333; font-family: arial; font-size: 12px; padding: 10px")
    )
)
map_osm.add_child(HOV)
map_osm.keep_in_front(HOV)

map_osm

Make this Notebook Trusted to load map: File -> Trust Notebook

From this map we can see that generally the most arrests happen near the city center, and decreases the farther away from downtown the neighborhood is. There are however some troublesome areas in the North West and North East quadrants of the city. The neighborhoods with a higher arrest count are also grouped together, which meets expectations.

This map would be helpful for knowing which areas to avoid if you want to avoid criminal activity.